Robustness is one of the most important issues in watermarking.In this paper

a blind image watermarking algorithm robust against affine transformation is proposed.This algorithm improves the robustness in three aspects:the embedding strategy

the watermark structure and the synchronization of the watermark.In the algorithm

the spread-spectrum-based watermark with a training sequence are embedded in the coeffcients of LL3 subband in DWT domain while a template is embedded in the middle frequency components in the DFT domain.In watermark extraction

we detect the template in the possibly corrupted watermarked image to obtain the parameters of geometric attacks and recover the image back to its original shape.Then we perform translation synchronization in the DWT domain by using the training sequence and extract the watermark.With the proposed algorithm

we hide a character string with 44 characters (264 bits) in a 512×512×8 image.Compared with other watermarking algorithms reported in the literature

the watermark thus generated is more robust

especially against geometric distortion

while having much higher bit rate.